高嘉轩,刘鸿瑾,施博,等.基于向量表的RISC-V处理器普通中断与NMI优化设计[J]. 微电子学与计算机,2024,41(4):112-122. doi: 10.19304/J.ISSN1000-7180.2023.0383
引用本文: 高嘉轩,刘鸿瑾,施博,等.基于向量表的RISC-V处理器普通中断与NMI优化设计[J]. 微电子学与计算机,2024,41(4):112-122. doi: 10.19304/J.ISSN1000-7180.2023.0383
GAO J X,LIU H J,SHI B,et al. Interrupt and NMI optimization in RISC-V processor based on vector table[J]. Microelectronics & Computer,2024,41(4):112-122. doi: 10.19304/J.ISSN1000-7180.2023.0383
Citation: GAO J X,LIU H J,SHI B,et al. Interrupt and NMI optimization in RISC-V processor based on vector table[J]. Microelectronics & Computer,2024,41(4):112-122. doi: 10.19304/J.ISSN1000-7180.2023.0383

基于向量表的RISC-V处理器普通中断与NMI优化设计

Interrupt and NMI optimization in RISC-V processor based on vector table

  • 摘要: 针对有实时性需求的精简指令集计算机(Reduced Instruction Set Computer, RISC)-V处理器中断响应延迟过长的问题,本文改进了中断响应中中断服务程序跳转地址计算的方式,扩展了不可屏蔽中断(Non-Maskable Interrupt, NMI)响应时的控制寄存器,提出了硬件矢量中断以及NMI相关控制寄存器扩展。硬件矢量中断提高了中断的响应速度,减少了中断响应的延迟。NMI扩展控制寄存器减少了NMI的响应延迟,减少了软件需要进行的保存现场操作。 利用VCS仿真验证了中断优化的正确性以及性能。 仿真结果表明,硬件矢量中断响应时间缩短了84.4%,响应速度提高为原本的6倍,NMI扩展控制寄存器减少了31个时钟周期的响应时间以及32个时钟周期的返回时间。

     

    Abstract: In response to the problem of long interrupt response delay in Reduced Instruction Set Computer (RISC) -V processors with real-time requirements, this paper improves the calculation method of the interrupt service program address in interrupt response, extends the Control and Status Register (CSR) during Non-Maskable Interrupt (NMI) response, and proposes hardware vector interrupt and NMI-related CSR extension. The hardware vector interrupt improves the interrupt response speed and reduces the interrupt response delay. The NMI extension control register reduces the response delay of NMI and reduces the need for software to save the context. The correctness and performance of interrupt optimization were verified using VCS simulation. The simulation results show that the response time of hardware vector interrupt is shortened by 84.4%, and the response speed has improved sixfold compared to the original. The NMI extension control register reduces the response time by 31 clock cycles and the return time by 32 clock cycles.

     

/

返回文章
返回